Role of Artificial Intelligence in the Toxicity Prediction of Drugs

Abstract

Therapeutic drugs are meant for treating the diseases, unfortunately, these drugs also possess unwanted adverse effects due to drug-drug interactions or drug-transporter interactions. Though the benefit-vs-risk ratio of the drug is well-thought-out before its approval; as a part of pharmacovigilance, drugs associated with life-threatening toxicities are withdrawn from the market. But, several patients get affected due to drug toxicities - even before its withdrawal. Hence, it is crucial to understand the drug toxicities in the early stage of drug development. Several animal species such as rat, rabbit, monkey and horse are used in preclinical safety studies, where the enormous number of animals and the time is consumed. Therefore, using artificial intelligence (AI) to predict the drug toxicity in early development stages could be a potential tool to address the said problem. Various AI tools are used for toxicity prediction which includes machine learning models – deep learning, neural networks, quantitative structure activity relationship, and molecular docking. A database with structural and physiochemical properties of the drugs along with its end-activity is used for the development of AI models to predict the potential toxicities of the drugs under development. In this chapter, we summarize the detailed use of different AI tools to predict the drug toxicities, in addition to the toxicity guidelines and a comprehensive overview about the existing AI models for predicting the organ specific toxicities.